29 Aug
2014
29 Aug
'14
4:54 p.m.
I agree with Alex and Jiri.
Here I have the same scenario, DNS round-robin and 2 dovecot instances. It works well, but depends on the client implementation.
I think keepalived or corosync/pacemaker is a more elegant solution too. I can help you implement.
On 08/29/2014 08:34 AM, Jiri Bourek wrote:
On 29.8.2014 13:00, Alex Crow wrote:
Hi Vijay,
No, this will not work properly. 50% of the time the client will still try to connect to the downed server.
You need something like keepalived or corosync/pacemaker.
Cheers
Alex
AFAIK some clients - with Thunderbird among them - will try the other IP address before reporting an error to the user.